A frontend veteran maps what to delegate to AI after six months

On November 6, 2026, a developer who says they have spent 8 years as a front-end engineer at a small company posted on V2EX under the title about handing work over to AI for the past half year. The company has no front-end team at all — just this one person, who owns three client platforms, builds, releases, and middle-of-the-night alerts. Over the past six months they say they have started giving every task that can be delegated to AI, and the post frames itself as practical notes rather than hype.

The tasks that genuinely work well, in their account, all share one feature: they have a standard answer. Regex and legacy compatibility code that used to cost half an hour and ten trial runs now take about two minutes, and after a quick check the result is usable. Last quarter, an old table page from about ten years ago needed an export-to-Excel feature, and AI wrote the whole xlsx section, which took only two style fixes before shipping. Minified error stack traces, pasted in, get roughly located and save about half the usual searching time. Writing comments, documentation, and technical-explanation documents for their boss is where the savings were largest.

The work that does not work out, they say, is exactly what fills their daily job. Whether a field can be deleted depends on a customer from three years ago whose data relied on it — information that lives in human memory, not in the code. The ugly parts of legacy code were written that way deliberately to work around some past problem, and an enthusiastic cleanup by AI means the site breaks in production. At release time, AI does not sign off. When the boss asks for something to look more premium, AI is just as confused as anyone.

At present they explicitly exclude four categories from AI work: payments, permissions, data deletion, and live hotfixes. They also describe an unease that goes beyond productivity: they used to deliver one page a week, which their boss considered normal, but now a colleague using AI can produce three proposals in a day, and the boss comes back asking why they are so slow. Their sense is that AI did not announce that anyone was out of the game; it simply raised everyone's expectations by one notch, and small companies feel that shift most acutely. The post was still in the discussion stage, with roughly 85 views, and the account reflects their personal account as of that date. Other developers' views and takeaways remain to be seen.